计算机技术与发展2017，Vol.27Issue(6)：41-45,50,6.DOI:10.3969/j.issn.1673-629X.2017.06.009
基于多叉树确定K值的动态K-means聚类算法
A Dynamic Clustering Algorithm of K-means Based onMulti-branches Tree for K-values
摘要
Abstract
K-means algorithm is the one of most classical clustering algorithms with repartition and has been widely used because it''s really concise and efficient.What''s more,it has advantages such as being easy to be implemented and low cost of complexity in running time and storing space.However,it''s normally initial number called K-value which cannot be precisely predicted by effective method.The initial clustering center used to be chosen randomly,so that the result usually converges to local optimal solution,which makes the latest clustering results inaccurate.The dynamic clustering algorithm of K-means based on multi-branches tree to determine the K-value is an improved one.The improved algorithm has been designed to determine the most reasonable K-value by dynamically dividing and merging cluster during the iterative process and partly solved the problem that clustering result converges to local optimal solution.Furthermore,exploration for corresponding data structure model has also been conducted to the investigation of the algorithm mentioned.Horizontal and vertical comparison with the binary K-means algorithm has been achieved.The comparison and analysis results show that the improved K-means algorithm is independent of improved global data sets,which makes it more suitable for distributed computing platform and that relative efficiency has been increased with increase of the size of the data set,especially in magnanimity data set.Therefore the improved K-means algorithm has promoted the clustering performance and can lead to a more stable clustering result.关键词
K-means/聚类/分裂/合并/多叉树Key words
K-means/clustering/dividing/merging/multi-branches tree分类
信息技术与安全科学引用本文复制引用
鲍黎明,黄刚..基于多叉树确定K值的动态K-means聚类算法[J].计算机技术与发展,2017,27(6):41-45,50,6.基金项目
国家自然科学基金资助项目(61171053) （61171053）
南京邮电大学基金(SG1107) （SG1107）
